Optimizing multi-layered networks towards a transparently optical internet

Abstract

Comparison between alternative technologies for switching and routing is difficult for several reasons: (i) the range of options is quite large and appears to demand knowledge over a very wide range -wider than we can reasonably expect to find in any one individual; (ii) in addition, each of the alternative technical solutions can potentially be used in combination with one of more of its competitors; (iii) the impact on the traffic being handled is difficult to model, and therefore it is almost inconceivable that we can predict the performance which can be delivered by complex configurations.In this paper we propose and develop a unified framework for multiple technologies that interwork together. To do this we adopt a simple model based on the principle that (i) technologies can be modelled as layers; (ii) traffic can be modelled as a Poisson stream of flows; and (iii) the dynamic (and static) choice between layers can be modelled by a strategy for splitting traffic between layers.Once this unified framework has been adopted, alternative technologies can be incorporated into in this model by choosing a collection of parameters which determine, for each layer, what it will cost to deliver a certain type of traffic and how traffic will be split between layers when there is a choice.

Layered design of mobile backhaul networks

“…In this particular example, the marginal cost of using a single STM-1 module over using multiple E1 modules occurs when it's required to have greater than eleven (11) x E1 modules in capacity in order to meet traffic load demands.…”

“…Definition of layered shortest-path routing Layered routing, as defined in [11], consists in treating the links at layer in n as the traffic at layer n-1. For example, if SDH channels are used, at a layer which sits above a transmission network with whole wavelengths, in a layered routing scheme of the sort investigated in this paper, the SDH channels would be regarded as the traffic to be carried on the wavelengths.…”

A new teletraffic approach for network planning and evolution prediction

“…For example, consider "mice" (the smallest flows), "kangaroos" (mid-sized flows), "elephants" (the largest flows), then the upper bound of the mice will be the lower bound of the kangaroos, and the upper bound of the kangaroos is the lower bound of the elephants [16]. Each class can represent a separate traffic stream that may be routed differently from other classes between the same OD pair and/or use different technologies.…”

“…These are derived using the parametric decomposition method (see [19] and references therein) considering end-to-end PPBP Traffic streams where PPBP processes merge and split in the network [16]. Then, having the statistics of the traffic on each link at any layer, we set link capacities using a queueing or loss model (depending on the relevant layer) according to GoS requirements.…”

“…We therefore focus first on an algorithm based on the shortest path principle which is prevalent in the Internet and has the key benefits of being scalable and distributed. This algorithm is a heuristic algorithm based on outsourcing demands from higher layers to lower layers, as in implementing a shortest path algorithm and capacity assignment in each layer and a fixed-point iterative algorithm [16].…”